Hacking Squeak With Apple MPW

agree at carltonfields.com agree at carltonfields.com
Mon Jun 7 21:17:44 UTC 1999


> One thing that I found with MrC (version 4.1.0a2c1) is that > at the highest
> optimization levels, it produces incorrect code for BitBlt, > which is why I
> don't use those opt levels.

I was likewise unable to get a working executable under the pre-release version 4 compilers.  On the other hand, the version 3 compilers work just fine and seem to generate superior performing code.

> Andrew, have you found out how to apply the VM patch for > MPW-generated VMs?
> MrC generates the same unnecessary bounds check for the > bytecode switch, only
> using LR instead of CTR for the branch. Throwing that out > might give you some
> of the bytecode level performance you're missing. I've > included my version of
> the patch method in case you need it.

I found that the existing patch code seems to do the trick with the version 3 optimized code, simply replacing "6" with "5".  A quick glance at Hans' code seems to indicate he did the same thing.  As I'm no PowerPC wizard, I'd be obliged if Hans would check my work and see if I'm missing something.





More information about the Squeak-dev mailing list